We study the decision-making of judges in an experimental setting resembling real world judicial decision-making. We gave 312 future judges 48 vignettes built from real data related to divorce cases involving children. We compared two different subject pools : judges who were asked to set child support awards with an advisory guideline and judges who were asked to set child support awards without any guidelines. We found that the introduction of such a guideline helps to reduce the disparity between judges (i.e., the variance in similar cases is lower when the subjects have the opportunity to use the guideline) but that this effect is not systematic, as an increase in heterogeneity was observed in some specific cases.